Re: HP Photosmart R032 problem

john allen pitney (pitney@students.uiuc.edu)
Tue, 16 Mar 1999 09:57:22 -0600

> > $ xscanimage [program starts, I choose /dev/scanner]
> > Preview Window [window appears]
> > [feed slide to PS, PS pulls in the slide a few mm]
> > Acquire Preview [preview acquired as slide feeds *into* the PS]
> > [preview looks okay, needs a lot of gamma correction 1.0 -> 2.2 or so]
> > [slide is now *inside* the scanner]
> > Scan [image is acquired as slide is pushed back *out* of the scanner]
> > [so far, so good]
> > Acquire Preview [again]
> > [this time, the preview is acquired and the slide returns to its
> > *out* position, unlike the first time]
> > Scan [this time, the slide feeds slowly back into the scanner, then
> > immediately reverses direction, and *then* data is acquired. Also, I
> > believe this is where the few mm of slip occurs.]
>
> This is normal behaviour for this scanner, it also happens with the
> HP windows software. Vuescan doesn't do a second pass over the slide
> (or strip) as it scans everything at once - perhaps that's why you
> haven't noticed this particular behaviour before. The stupidity seems
> to be internal to the scanner.

OK. I haven't noticed it, but then I don't think that I've ever caused the
HP software to reacquire a preview, since color correction shows up in its
preview window. I was re-running the Acquire Preview to see the effect
of editing the gamma curve in xscanimage.

> > Also, my test slide is a fairly typical scene, albeit without any sky
> > showing. I find that the default straight gamma curve results in a very
> > dark image--to fix it with gimp's levels tool, I need to set the white
> > point to 153 (out of 255) and the gamma to about 2.50. Of course, the
> > image gets reduced to something like 5 bits of precision after that kind
> > of rough treatment. Does this mean the R032 is going to require a
> > downloaded tone map just to produce a decent image?
>
> For dark slides, increase the Exposure Time to for example 300% and
> apply a gamma of 2.2 if you have a PC monitor.

This slide isn't particularly dark, but the image I get with a gamma of
1.0 is simply unusable. A histogram shows that it has almost no
pixels with a value above about 150 (out of 255). This is with the
recommended 150% exposure. Applying a gamma and white point correction
to the scanned image does fix it up, but, as I pointed out, it leaves
an image with very many large gaps in its histogram and markedly reduced
color resolution.

Is it the right thing to do, to acquire an 8-bit/channel image with a
flat 1.0 gamma curve, then apply a gamma of 2.2 to the result?

> About tonemaps, I feel it is better to do this in software (on the
> frontend side) instead of implementing it for every backend.
> Adding some intelligence or densitometric data to the frontend will
> benefit all scanners and not just the Photosmart.

I agree there--the densitometric data should be in the frontend. My
concern is that possible the HP PS R032 can't output a usable 24-bit
image unless there is a (10x8?) tone map downloaded to it (which seems to
be happening in the VueSmart logs).

> > Last night, I tried looking at a log file produced by VueSmart. I put
> > some up at the following URLs:
>
> I don't know of any interesting SCL options that this scanner could
> still have and that's not incorporated into the newest backend
> already. What is it that you are looking for in the vuesmart logs that
> you like to have?

I just wanted to see how VueSmart was doing such a good job! So far, I
haven't managed to get a comparable image from xscanimage/hp backend. Doing
so will require either manipulation of 10-bit data in software or
downloading a (10x8?) tone map to the scanner.

By the way, what does the 10x8 notation mean? Map 10-bit data to 8-bit?

Thanks for the reply!

John

--
Source code, list archive, and docs: http://www.mostang.com/sane/
To unsubscribe: echo unsubscribe sane-devel | mail majordomo@mostang.com